The MARVEL Data Corpus: A Secure and Privacy-Aware Multimodal Big Data Repository for Smart Cities Research

Smart cities increasingly rely on Big Data, IoT, cloud computing, and AI to enhance urban services. However, limited access to public datasets hampers research and innovation. This paper presents the MARVEL Data Corpus, a large-scale, multimodal repository developed under the EU-funded MARVEL project. It includes 1.1PB of anonymized audio-visual data from Malta, Gozo, Trento, and Novi Sad, captured via surveillance systems and drones. The repository features 52,000 snippet files across 64 datasets, with 35 publicly available under a Creative Commons license. Use cases cover traffic monitoring, emergency response, public safety, and event analysis. Data is anonymized at the edge to ensure privacy compliance. The infrastructure is built on Hadoop and HBase, with access via APIs and a web-based Angular GUI. An Augmentation Engine enables data transformations to support AI model training. The MARVEL Data Corpus addresses the lack of open, privacy-aware smart city datasets, promoting AI research and data-driven innovation.